Broken — Actors

Broken actors are not edge cases but inevitable in sufficiently complex agent systems. As we deploy more autonomous agents in open-ended environments, we must move beyond simple fail-stop models. Future work should focus on graceful degradation protocols and real-time brokenness detection without assuming perfect knowledge of the true objective.

A satisfies three criteria:

A configurable policy engine that executes when an actor enters a Broken state. broken actors

In the context of the Ocarina of Time & Majora's Mask (OoTMM) Combo Randomizer Broken actors are not edge cases but inevitable

As artificial intelligence systems transition from single-model deployments to complex multi-agent architectures, the concept of the “broken actor” becomes critical. A broken actor refers to an agent whose behavior deviates from its design specifications due to internal malfunction, reward misspecification, or environmental interaction. This paper defines broken actors, categorizes their failure modes, and discusses implications for robustness, safety, and alignment. A satisfies three criteria: A configurable policy engine